Macro segregation in continuous cast HSLA steels

(1)

EXAMENSARBETE INOM TEKNIK, GRUNDNIVÅ, 15 HP

STOCKHOLM, SVERIGE 2019

Macro segregation in

continuous cast HSLA steels

W ith correlation to impact toughness AUGUST ÅSTRÖM

MORGAN STEN

KTH

SKOLAN FÖR INDUSTRIELL TEKNIK OCH MANAGEMENT

(2)

Abstract

The report reviews macro segregations in continuous cast steels and possible correlations to impact toughness. The purpose of the thesis is to investigate centreline segregates and V-segregates to see which of them that affect impact toughness the most. Apart from a literature study, Charpy-V data was collected with permission from SSAB for two steel types, grade A and grade B, each with respective dominant segregation type. The collected data was yielded in three individual Charpy-V tests at different spots on the metal sheet, derived within a close area. The average value of these tests were used in statistical analysis to observe the spread of values in different heats of the two steels. Additionally, the specimens were etched and captured in cross-section. Results indicated that for the data of grade A, where centreline segregates were dominant, the spread of values was higher than for the data of grade B. The conclusion is that centreline segregations are worse in relation to impact toughness, since higher deviations translates to less predictable properties from a customers perspective.

(3)

Sammanfattning

Denna rapport granskar makrosegringar i stränggjutet stål och eventuella kopplingar till slagseghet.

Centrumsegringar och V-segringar undersöks för att undersöka vilken segringstyp som har störst effekt på slagseghet, vilket är syftet med denna avhandling. Förutom en litteraturstudie, hämtades Charpy-V data med SSAB’s medgivande från två olika stålsorter, kvalitet A och kvalitet B, med respektive dominant segringstyp. Datan som inhämtades erhölls från tre stycken Charpy-V tester för varje plåt, från ett närliggande område. Medelvärdet för dessa datapunkter användes i statistisk analys för att obeservera spridningen av datapunkter i olika charger av de två stålen. Dessutom, erhölls segringsbilder för respektive slab från SSAB. Resultaten visade att datapunkterna för kvalitet A, som hade centrumsegring som domiant segringstyp, var mer spridd än datan för kvalitet B. Således, är slutsatsen att centrumsegring är värre i relation till slagseghet eftersom en högre avvikelse leder till mindre förutsägbara egenskaper från en kunds perspektiv.

(4)

1. Introduction

In this chapter, background to this work and its purpose will be established. In addition, ethical and social aspects will be reviewed.

1.1. Background

Macro segregation is a phenomenon in alloyed metals that emerges during solidification, causing variations in composition depending on where in 3D-space the slab is observed. The hypothesis is that the gradient in composition will negatively affect mechanical properties. On the continuous casting site at SSAB Oxelösund, it is already known how different types of macro segregations form. However, what has not yet been studied is how these segregations affect the mechanical properties of the end- product. This study will mainly treat two types of macro segregations: centreline segregation and V- segregation. Therefore, two high strength low alloyed (HSLA) steels will be studied, each with respective dominant segregation-type. The mechanical property subject for testing in this report is impact toughness.

1.2. Ethical and social aspects

Since this report features macro segregation in continuous cast steels and their effects on properties, there is no direct relationship to ethical aspects. If the results from this report proves to be applicable for every case of determining which segregation is the worst, the company can use that knowledge in their design. That would then lead to the fact that the products made from their steels would have a longer life cycle and, in that regard, it would be beneficial for the environment. However, that assumption is far-fetched which has affected our decision not to include discussion related to ethical and social aspects.

1.3. Objective

The aim of this thesis project is to determine what segregation type is the worst in relation to impact toughness, thereby helping SSAB decide what segregation to prioritize removal of. To achieve this, data from processes will be analyzed in addition to mechanical testing. Since the phenomenon of macro segregation is complex, a literature study will first be conducted. Following, the group will make a field trip to SSAB and retrieve data to analyze, thereby hopefully reaching a conclusion to what segregation is the worst. In addition, a discussion to future improvements will be made.

(7)

2

2. Variables and phenomena related to segregation

In this chapter, different phenomena occurring during solidification in the process of continuous casting will be treated, including the phenomena of macro segregation. Furthermore, the dominant mechanism of formation to different types of segregation will be presented. Since segregation depends on several phenomena such as solidification shrinkage and crystal growth to name a few, those will also be explained further. Additionally, since there are different types of crystal structures present in a typical slab viewed in cross-section, two common crystal structures will be examined.

2.1. The phenomenon of macro segregation

In the 20th century, Scheil derived equations modelling the phenomenon of micro segregation. On the micro-scale, assumptions made during derivation make a decent approximation. On the macro- scale however, most of the assumptions are simply not realistic to make [1]. An example is the assumption that the melt and solid share the same molar volume, which implies that they also have the same density. In reality, a solidifying melt will increase in density resulting in a decrease in its molar volume. Since the pressure inside the melt is practically zero, melt is sucked through the dendrite network. When this happens, local deviations in composition appear; a phenomenon known as macro segregation.

The effects of macro segregation are especially important in large castings for two reasons. First, solutes in the solid state have low diffusivity and second, macro segregations cover a wide distance.

These effects make it difficult to remove deviations in composition with the traditional methods used for micro-scale segregations, which is usually some kind of heat treatment [2]. Other factors that influences the amount of macro segregation in continuous cast steel include thermal contraction, solidification shrinkage, natural convection and equiaxed crystal growth. [3]

2.2. Solidification of molten steel

In the late 19th century Russian metallurgist Tschernoff published research describing three distinct zones of microstructures in cast steels: a surface zone with mostly small equiaxed crystals, a middle zone with columnar crystals and an inner zone with equiaxed crystals, illustrated in Figure 1 [4]. These different kinds of microstructures and their respective formation will be discussed later in this chapter.

Although Tschersnoff’s report originally covered ingot casting, the research has been adapted as a good model in most types of large-scale castings, including continuous casting.

(8)

3

Figure 1- Schematic structure of cast HSLA steel [4]

Another, Swedish metallurgist called Hultgren found that by varying temperature during solidification it is possible to manipulate the formed microstructures [4]. It was particularly shown that an increase in casting temperature leads to an increase of the columnar zone, thereby decreasing the central zone.

Consequently, a decrease in casting temperature leads to a decrease in columnar crystals, or in extreme cases even the absence of them. Hultgren also found that stirring the melt during solidification influences the formed microstructures. In addition to the research it has been confirmed by experiments that the surface zone is small, while the size of the middle zone containing columnar crystals and the inner zone with equiaxed crystals vary considerably depending on factors that will be further discussed below.

2.2.1. Equiaxed crystal growth

The small crystals in a cast steels are mostly present in two zones: the surface zone (sometimes referred to as the chill zone) and the inner zone. [4]

The surface zone is the outer layer of a casting and where the first solidification takes place. In the process of continuous casting, water cooling is present from the mould to the horizontal zone.

Consequently, the first melt that solidifies does so rapidly by reaching a critical temperature of nucleation. The water-cooling system accelerates this process as the chilled surfaces are in direct contact with the outer layer of the cast, hence both conduction and convection can take place.

Furthermore, since temperature changes fast in the surface zone, there will be multiple nucleation sites which ultimately leads to small crystals.

The inner zone is the centre of the strand. Since the melt initially starts to solidify in the outer layer, the inner zone is inevitably where the last solidification takes place. After the columnar crystals have formed, they will continue to affect the flow of heat. Potential latent heat gets transported away, leading to quick cooling in the inner zone. The inner zone is not cooled as fast as the surface zone, where direct heat conduction between the melt and mould/nozzle water plays a big role. Still, it is fast enough to favour nucleation of new crystals over crystal growth.

(9)

4

2.2.2. Columnar crystal growth

When the small crystals in the surface zone have formed, that layer will block the possibility of direct contact between the melt and the mould wall or water streams from the nozzles, leading to slower transportation of heat from the strand into the cooling water; the heat has to pass through a layer of solid metal that has lower thermal conductivity than the melt. Also, the distance itself between the melt and the water has increased because of the surface layer. These factors lead to a cooling rate slow enough to benefit few nucleation points and growth in the inner zone. However, there already exists a solid area which is easier to nucleate on. Therefore, the nuclei in the second solidification zone grows as columnar crystals. Columnar crystals are characterized by the fact that their growth direction is parallel to the temperature gradient. Each individual crystal consists of several parallel dendrite arms which all reach approximately equal distance into the melt.

2.2.3. Cooling shrinkage

Since cooling shrinkage is one of the factors influencing the grade of macro segregation, it is an important phenomenon to grasp. The cooling shrinkage does as previous stated, create cavities that indirectly induces composition deviations when residual enriched solutes gets sucked through the mushy zone [5]. The dominant mechanic responsible for cooling shrinkage is the deviant density of molten metal and its solid counterpart; i.e. metals are less dense in its liquid state [6]. Therefore, when this phenomenon occurs during solidification, it causes local increases in density, leading to a decrease in volume. One cubic metre of molten steel weigh approximately 6.9 tonnes while one cubic metre of solid steel weighs in at about 7.9 tonnes. Calculated per tonne steel the solidification shrinkage amounts to 18 litres [7]. Furthermore, since alloys solidify over a temperature range, solidification energy is gradually released into the melt. The temperature in the centre of the strand can be approximated as constant, but when no further energy is being released from the surface layer, the rest of the melt solidifies fast in the moment that the two solidification fronts meet, leading to shrinkage. In continuous casting, it is common to use water cooling for a smoother solidification process [1]. Apart from traditional cooling shrinkage during solidification, there exists other types of shrinkage, e.g. shrinkage of the liquid, and patternmaker’s shrinkage [6]. These two however, are often dismissed; in the continuous casting process the shrinkage of liquid is eliminated when new molten metal is inserted continuously from the top, and the patternmaker’s shrinkage is handled through manipulation of alloy composition.

2.3. Centreline segregation

In the 1980’s, an experiment was performed where steel was continuously cast with a rate between 2.5-3.5 m/min [1]. Several tests with varying carbon concentration showed that the amount of macro segregation was heavily dependent on the cooling rate, and not so much on the amount of carbon.

Furthermore, FeS was injected with a nail into the strand. With the help of sulphur prints, the flow of molten steel could be mapped, and with it, the formation of a pipe below the nail. When the two solidification fronts met locally, a bridge was formed on points that were exposed to lower temperatures or in some other way were more beneficial for growth. Solidification shrinkage and solubility of alloy elements are the main parameters in this type of segregation.

(10)

5

When bridges form, they block the liquid from flowing further in the strand [8]. When the permeability of melt is too low, the load from the accumulated melt deforms the bridge into a U-shape before it breaks. This results in that “the residual inter-dendritic melt is squeezed out in this plastic deformation process “ [8].

When the melt transforms into solid metal, the atoms get more arranged; i.e. the entropy decreases due to the transformation of an amorphous liquid into a crystalline structure. These lattices take up less volume than an amorphous melt and therefore solidification shrinkage appears [9]. When this phenomenon occurs in the strand, the melt at the higher altitude compensates; in other words, due to the law of least resistance, enriched melt that has a high concentration of solute gets sucked into the dendritic network and fills the cavities. The melt gets enriched because the dendrites in the solidification front grows with micro segregation and some of the solutes flows to the melt. To minimize segregation it is common to use the process of soft reduction, which will be reviewed in chapter 2.5.1. Figure 2 shows what a slab with centreline segregation looks like after etching.

Figure 2 - A typical centreline segregate after etching [5]

(11)

6

2.4. V-segregation

In addition to the formation of centreline segregations, it is common for the material to form V- segregates. These segregates come in two different kinds: simple V-segregates, and closely packed V- segregates. The formation of closely packed V-segregates is favoured in the angle of ~45 degrees respective to the central axis (see Figure 3). Furthermore, closely packed V-segregates are more likely to form in an area with wide equiaxed zone, and the smaller the crystals are in size, the more abundant the V-segregates become. As previously stated, in the continuous casting process, these conditions are commonly created in the later stages of solidification. The other type of V-segregates, simple V- segregates, are formed in the same way, but are more plentiful when the columnar zone is more widespread than the equiaxed zone. It is common for bridge-formation to take place beneath the simple V-segregates. [1]

In general, the knowledge of V-segregation is low as they have not been subject for extensive study.

A few studies that have treated the mechanism of formation of V-segregates have been similar in the regard that V-segregates are thought to be formed by flow of interdendritic liquid induced by a negative pressure gradient, which in itself exists as an effect of cooling shrinkage [10] [11]. In continuous casting, V-segregates can form when there is a wide equiaxed zone containing small crystals. The wideness of the equiaxed zone affect how close packed the V-segregates are. With a slim equiaxed zone, the columnar growth is favoured and the distance between the segregates, the V- shapes, increases. One theory behind the cause of this is that the solidification transformation in the strand usually is not in equilibrium; the element with highest solubility solidifies first, pushing away the excess solute to the centre. If there are small equiaxed crystals present when external pressure is applied in the mechanical soft reduction process, the solutes from the centre can move in between the crystals with the enriched melt, sometimes resulting in the formation of interdendritic structures.

Figure 3 – V-segregates in a continuous cast slab cross section [1]

(12)

7

2.5. Solidification in the continuous casting process

Figure 4 shows a typical cross-section of a continuous cast slab. As described in chapter 2.2, the first and utmost material (the surface zone) is mostly made up of equiaxed crystals. This is illustrated in the picture near the sides, where macrostructure seems to be smooth. The surface zone is then followed by the columnar zone, and in the very centre, the equiaxed zone. The dark line in the centre of the picture illustrates the phenomena of centreline segregation. It is important to take into consideration that in reality; the material will solidify in three dimensions. The reason that segregates are concentrated to the centre is explained in the same way that micro segregation is described; i.e.

the last melt solidifies in the centre. The concentration of segregates is further increased in the process of soft reduction, a process in the end of the strand which will be discussed later in chapter 2.5.1.

Figure 4 – A typical macrostructure of a continuous cast slab, shown in cross section [4]

The process of continuous casting features different kind of segregates, where the most common and well-studied are centreline segregations. According to Lesoult [12], centreline segregates are also the most harmful types of segregates and regardless what the end product might be, these types of positive segregates must be kept to a minimum. Although the method of soft reduction has a main purpose to avoid the return of solutes into the melt by forcing them to stay in the solidification front and solidify, it also helps to reduce the likelihood of hollow spaces and pores forming inside the material. Since the melt solidifies partially during extraction there is clearly a point in the production- chain where the last melt inside the strand solidifies. Because the strand is constantly moving forward during this process a cone-shaped pipe of the last melt is formed. In continuous casting however, this is not a widespread problem since new melt in constantly being fed into the strand. The only part affected is therefore the last part of a heat, which is cut of and recycled.

(13)

8

To reduce the effects of centreline segregates in practice, it is common to steer the structure towards some equiaxed zones as they reduce refining action [13]. The equiaxed zones also help prevent the phenomena of bridge formation that would otherwise appear in other structures such as columnar crystals. Bridge-formation is a consequence of a solidifying front that sometimes partially cuts off the rest of the melt that by cooling shrinkages will lead to the formation of pores and hollow spaces. These are obviously not good for neither final property of the produced material. However, the consequence of using equiaxed zones is that they can form their own partial bridges and according to Ghosh [13], it is essential to keep the equiaxed crystals small to reduce this effect.

In another extensive study on solidification in continuous casting it was found that cavity volume in the strand primarily depends on chemical composition, and the solidification mechanism of the steel.

Furthermore, steels with a high concentration of alloy elements, i.e. carbon, have a greater tendency to form pores than steels with a low concentration of alloy elements [14]. The study also concluded that cavities formed in equiaxed strands generally are smaller by volume than in strands with dominance of columnar crystals.

2.5.1 Soft reduction

The enrichment of alloy elements in the centre decreases the quality and tends to increase the risk of cracking. To minimize the amount of segregation, external pressure from rolls in the horizontal zone can be used in a process referred to as mechanical soft reduction. Since ferrostatic pressure is directed from the strand towards the rolls, the melt wants to flow in the same direction. By adding pressure from the support rolls, the melt can be forced to hold its position and solidify. This is because the process of soft reduction compensates for both shrinkage during solidification by moving the edges closer, and the ferrostatic pressure. However, it is crucial to know where the solidifying front is located. This is what is referred to as “metallurgical length” and can be seen in Figure 5. If any other part of the strand is object to the applied pressure, the process is completely useless and will only cause deformation on the rolls. Therefore, it is important to know the metallurgical length [9]. There exist other reduction processes but this one is the most effective and is what SSAB use.

Figure 5 – A schematic illustration of where soft reduction is applied [15]

(14)

9

2.6 Effects of segregation to material properties

When the material has been casted into slabs it undergoes further processing to improve properties.

One common process is hot rolling, which is a form of mechanical deformation during warm temperatures. Despite most of the material undergoing plastic deformation and recrystallization, the specimen’s macro segregations will remain close to unchanged. Therefore, it is important that desired macrostructure of the cast is finalized in the casting process, since that is where most alterations can be made [16].

In general, the connection between macro-scale segregation and material properties is poorly researched and it has been proven difficult to even find papers treating our specific conditions. With that in mind, research published by Jernkontoret has found that the effects of segregates are mostly the same in micro- and macro structures; the main difference of course being that macro-segregates are more severe since the scale is greater [7]. Furthermore, the effects of macro segregation cannot be removed through traditional heat treatment as distance for diffusion is vastly too large.

On the microscale, a segregation in the grain boundary can lead to fracture due to local brittleness [17].The same seems to be true for macro-scale segregates; local deviations in composition is partly what causes the centreline segregation. The impurities in the centre of the strand is later indirectly inducing the hollow spaces and widespread of bridge-formation.

(15)

10

3. Method

On March 18th-19th the group conducted a field trip to SSAB AB in Oxelösund. The purpose of the trip was to get a better understanding of the generally complicated process from preparation of raw material to cast slab. Even after the cast, the process is continued with hot rolling and mechanical testing.

3.1. Experimental material

In specific, two specimens with different compositions were selected to analyse and will be referred to as grade A and grade B respectively. Due to organizational secrecy however, exact alloy compositions and parameters of processing remain a blind spot in order not to compromise SSAB business. The two specimens were equally tested, and data was collected, but again it must be emphasized the importance that most of these data are to be treated with utmost secrecy. Therefore, only a handful of data will be present later in the results.

The experimental study was done by first plotting all the data given by SSAB to see the distribution of energy to specimen thickness. This was done to see the spread of values for the heats to choose specific heats to analyse further. After that, specific values of specimen thickness that had a wide range of impact toughness were chosen; the thought is that these could be likely to contain abundant macro segregation because of the fact that different elements affect the impact toughness. Following, pictures of their respective slab after etching were observed along with their centrum quality index, determined visually by operators at SSAB. The pictures of the steel grades respective slab gave the possibility to include their segregation types. The visual inspection method involves a comparison with reference pictures of varying degrees of segregation. 𝐶𝐶𝐶𝐶 = 0 is the optimal centrum quality and anything above 𝐶𝐶𝐶𝐶 > 0 is considered to contain some unwanted properties such as tendencies of macro segregation. There is no definite upper limit for the scale. This then gave the possibility to connect the segregation images to the results in impact toughness. The etchant used on the slabs for the segregation pictures dissolves elements that are sensitive to it. The thicknesses chosen to look further into were 12 mm and 15 mm for both steels. Furthermore, the group analysed the data statistically in accordance to formulas in chapter 3.3.

Finally it should be noted that details regarding the used processes were unknown to the group, including for example the type of etchant used, how the material was etched, how it was later photographed and other possible pre-treatment.

(16)

11

3.2. Charpy-V impact test

Charpy-V testing is a standardised method to measure a materials energy absorption during impact fracture. The measurement device consists of a pendulum and a hammer of which is dropped from a predetermined height onto the standardised sample [18]. Depending on the measured height of the pendulum on the other side after impact, an impact absorption energy can be calculated with the help of the traditional energy principle. This can later be related to other material properties and will be furthered discussed in the forthcoming chapters. This is primarily how our specimen will be tested and the energy absorbed will be referred to as EMV, energy mean value; derived as an average value of E1, E2 and E3. These are individual tests taken from different regions on the sheet and represent the energy absorbed of the specimen. The temperature in the tests was set to -40 ֯C. A general setup for Charpy-V testing can be observed in Figure 6.

Figure 6 – Schematic illustration of a typical Charpy-V notch test [19]

(17)

12

3.3. Statistical analysis

In order to analyze results in later chapters, two formulas need to be introduced. Firstly, sample standard deviation [20] can be calculated using Equation 1 below.

𝑠𝑠 = � ∑ (𝑥𝑥^𝑛𝑛_𝑖𝑖 _𝑖𝑖− 𝑥𝑥̅)² 𝑛𝑛 − 1

(1)

Where 𝑛𝑛 is sample size, 𝑥𝑥𝑖𝑖 the individual data points, and 𝑥𝑥̅ the mean value. The equation is primarily used as a dispersion measurement. If values in a dataset 𝑛𝑛 are close together, (𝑥𝑥𝑖𝑖− 𝑥𝑥̅) will be small, leading to a low value of 𝑠𝑠.

Following, in order to estimate correlation between two variables in a dataset, the Pearson correlation coefficient can be used [20].

𝑟𝑟 = ∑ (𝑥𝑥^𝑛𝑛_𝑖𝑖 _𝑖𝑖− 𝑥𝑥̅)(𝑦𝑦_𝑖𝑖− 𝑦𝑦�)

�∑ (𝑥𝑥^𝑛𝑛_𝑖𝑖 𝑖𝑖− 𝑥𝑥̅)²�∑ (𝑦𝑦^𝑛𝑛_𝑖𝑖 𝑖𝑖− 𝑦𝑦�)² (2) Where 𝑛𝑛 is sample size, 𝑥𝑥𝑖𝑖 and 𝑦𝑦𝑖𝑖 are the individual data points, 𝑥𝑥̅ and 𝑦𝑦� the mean value for each sample. A perfect linear dataset in the positive direction results in value of 𝑟𝑟 = 1. Following, when 𝑟𝑟 = −1 it implies a perfectly negative linear fit. When 𝑟𝑟 = 0, it implies no correlation ever between datapoints in 𝑋𝑋 and 𝑌𝑌.

(18)

13

4. Results

As previously stated in chapter 3, the two specimens were mechanically tested, and the results are shown below for each of the steel types.

4.1. Specimen of HSLA grade A

In total for all heats, data for specimen of type grade A is shown in Figure 7. On the Y-axis, EMV-value is plotted with respect to specimen thickness. The chart contains all the heats available to the group for grade A. A few of these heats were selected for further investigation and will be referred to as H1, H2, H3 and H4 respectively.

Table 1 on the following page shows the result for each selected heat of grade A. Since specimen thickness varies along a heat, selection was made in order to maximise a certain thickness’s representation. Furthermore, in order to examine possible connections between mechanical properties (here represented by EMV value) and degree of macro segregation, one heat with low EMV- values, and one with high, was selected. The respective heats for grade A are shown on the next page in Table 1. Data points represents the number of samples in the specified thickness, while total data points refer to all samples in the heat.

Figure 7 – Sample spread of all grade A heats

(19)

14

Following, the heats of grade A were all etched and captured in cross-section shown in Figures 8-11 below. A distinct line can be distinguished by the naked eye, widespread over the centre of the specimen.

Figure 8 – Heat H1 Figure 9 – Heat H2 Figure 10 – Heat H3 Figure 11 – Heat H4

The mean sample standard deviation was calculated regarding E1, E2 and E3. These are as previously stated the individual measurements for each sample in the heat. The mean is the average of these, and in a nutshell, indicates how accurate measurements are regarding sample size. Data is available in Table 2 below.

Table 1 – Data from heats H1-H4

Heat Thickness [mm]

Low EMV High EMV Average EMV

Data points

Total data points

H1 12 46 55 48 2 4

H2 12 19,3 26 22,1 4 10

H3 15 22 30,7 30,2 2 7

H4 15 33 34,7 33,8 2 14

(20)

15

Lastly, the histogram shown below in Figure 12, is a chart illustrating how many specimens were found in a specific EMV range. Furthermore, it is overlaid by a normal distribution curve to scale, purely for visual interpretation.

Figure 12 – Illustration of EMV distribution across heats H1-H4 Table 2 – Additional data subject for statistical analysis for heats H1-H4

Furthermore, some additional data which might be relatable to macro segregation were calculated using equation X. These are shown in Table X below. Heat Average E1-E3

𝑠𝑠

𝐶𝐶𝐶𝐶 EMV/Thickness 𝑟𝑟

H1 ±16,0 1 0,27

H2 ±6,8 0,5 0,88

H3 ±9,4 1 0,33

H4 ±10,3 1 0,26

(21)

16

4.2. Specimen of HSLA grade B

In total, the data for all heats and specimen of grade B is shown in Figure 13. On the Y-axis, specimen EMV-value is plotted with respect to specimen thickness. In general, the described methods applied to heats of grade A was used for grade B as well.

At first glance, the sample size for type grade B is huge in comparison to grade A. What unwanted effects this might have on results will be later discussed in chapter 5. Furthermore, the selection of heats is shown in Table 3, referred to as S1, S2, S3 and S4 respectively.

Table 3 – Data from heats S1-S4

Heat Thickness [mm]

Low EMV High EMV Average EMV

Data points

Total data points

S1 12 76 172 115,1 18 85

S2 12 50 62 56,7 23 57

S3 15 55 72 63,2 11 67

S4 15 78 98 89,4 12 64

Figure 13 – Sample spread of all grade B Heats

(22)

17

The corresponding heats captured in cross section of grade B can be viewed in Figure 14-17. Compared to grade A there is no clear centreline. The microstructure looks more homogenous at first glance.

Figure 14 – Heat S1 Figure 15 – Heat S2 Figure 16 – Heat S3 Figure 17 – Heat S4

Furthermore, Table 4 shows the calculated 𝑠𝑠- and 𝑟𝑟-values for each heat. The calculated 𝑠𝑠-value is an average of the deviations for each individual sample in respective thickness. Correlation on the other hand is calculated for all samples in the heat regardless of thickness.

Table 4 - Additional data subject for statistical analysis for heats S1-S4

Heat Average E1-E3

𝑠𝑠

𝐶𝐶𝐶𝐶 EMV/Thickness 𝑟𝑟

S1 ±32,2 0 0,76

S2 ±7,3 0,5 0,84

S3 ±8,9 0 0,55

S4 ±21,0 0 0,80

(23)

18

Finally, Figure 18 illustrates the distribution of samples in heats S1-S4. Since sample size is quite big in these heats, it is quite clear that the EMV is not properly distributed in accordance with normal distribution, but nevertheless it is a clean comparison.

Figure 18 - Illustration of EMV distribution across heats S1-S4

(24)

19

5. Discussion

The fact that research related to V-segregates is lacking in most areas is likely a consequence to their rather complex formation and nature. Since formation occurs inside the strand during solidification, V-segregates are likely hard to monitor in practice. Furthermore, modelling of the phenomena is viewed as problematic as a wide variety of factors could be influencing their formation and because the casting process must be “closed” in order to keep elements such as oxygen away. Like stated in chapter 2.4, V-segregates are particularly likely to form where there exists a wide equiaxed zone with small crystals, influenced by factors such as cooling shrinkage and ferrostatic pressure. Figure 14-17 show that neither specimen of grade B has a clear centreline. Instead, the centre along with the rest of the sample can be interpret as homogeneous; at least the irregularities seems to be consistent. This would correlate to a wide equiaxed zone as those crystals are randomly oriented. In contrast, the columnar crystals seem to be growing mostly in a specified direction or at least with predetermined orientation, which should make identification easy in the specimen captures. Not many of these columnar crystals are found in the captures of heats S1-S4. It is therefore rational to assume that the specimens of grade B all contain a wide zone of equiaxed crystals. One fact to consider is of course that etching affecting the specimen’s surfaces can alter what later is interpret by the human eye.

5.1. Analysis of segregation captures

Contrary to the specimens of grade B, Figure 8-11 illustrates a clear centreline in all specimens of grade A. Furthermore, the amount of columnar crystals seems to be more widespread and are easy to spot.

In the captures, the columnar crystals are shown as the locally uniform sections starting close to the edge, with growth inwards towards the centre. In relation to the theory of solidification and crystal growth, there exists a clear area where growth is believed to go from columnar to equiaxed.

Additionally, captures of heats H1-H4 show that a crack has formed as a line in the middle. These cracks are likely the consequence of centreline segregates which are further illuminated by the process of etching; the enrichments in the steel are sensitive to the etchant used. Therefore, the material in the centreline has been etched away, leaving only a crack in its place.

5.2. Segregation and impact toughness

As stated before, there is not a lot of research on the connections between segregations and the mechanical properties of the end-product. One reason could be that it is hard to isolate exactly what property changes and moreover test results are direct products of the segregation without involvement of other defects. It is not possible that there would be no connection at all between the degree of segregation, the type of segregation and mechanical properties because it is well known that different alloys have different properties and in a segregated material, there are composition gradients in the material.

(25)

20

5.3. Correlation between specimen thickness and impact toughness

What can be seen in the results is that the impact toughness of the steels increases with thicker specimen. This is a product of a bigger cross-section area, which means that the impact energy is distributed to a larger amount of material. It is hard to determine the exact relationship between thickness and energy absorption. However, the correlation between sample thickness and impact toughness can be investigated using respective 𝑟𝑟-values for each heat. At first glance, it is clear an increase in thickness directly leads to a higher impact toughness; all cases in this report (heats H1-H4, S1-S4) returned a 𝑟𝑟-value greater than zero. This was expected as it is not hard to imagine that a thicker material sample will be able to absorb more energy. Regarding what material is most affected by an increase in thickness between grade A and grade B, it can be said that heats S1-S4 in general seems to have had a higher value of 𝑟𝑟. For future research, it could be interesting to look at what elements affect segregation the most, and what their correlation are to impact toughness.

5.4. Low and high EMV

The results regarding heats H1 and H2 show that the specimens with worse 𝐶𝐶𝐶𝐶 has higher impact toughness. This could be contradictory because it is easy to think that a high degree of segregation would return a low impact strength. However, if the material in the specimen was for example enriched by an element that has a higher impact toughness than the base material, of course the impact toughness will be higher in that local area. So rather than making this connection it is more accurate to view the actual variation of impact toughness as an issue because if one part of the material is enriched, it will lead to another part being depleted. Resulting in that the product has inhomogeneous properties. This is why the spread in the data was in focus.

5.5. Heat H2 versus heat S2

Moreover, additional comparison can be made between H2 and S2. The selected tests from these heats have the same thickness and same 𝐶𝐶𝐶𝐶 value. However, if we regard the standard deviation of the specimens that have been derived from different parts of the plate sheet, we see that S2 on average varies more from its individual measurements. This could indicate that the segregation in S2 affect its EMV more than H2 affect its EMV since the other factors are the same. That would then mean that V-segregation, which is the dominant type in grade B, has a bigger influence on impact toughness. However, this conclusion is affected by the fact that for H2, there are a lot less data points for 12 mm than for S2, but at the same time the fraction of data points with this thickness compared to all data points in the heat are the same for H2 and S2. This conflicts with the literature study where it was stated that centreline segregation is the worst for any material in any application. Furthermore, it is a fact that in the grade A samples, all the other slabs observed have a higher 𝐶𝐶𝐶𝐶 and the opposite is true for the grade B samples. In relation to this, the H2 and S2 have the lowest standard deviation amongst their population. Another thing that contradicts this theory to be an accurate conclusion is that the standard deviation for H2 is 6,8 and for S2 it is 7,3 and that means that the difference between them is 0,5 which is small.

(26)

21

5.6. Spread of data versus segregation

In Figure 13 and Figure 19 it is shown that the data of grade A is somewhat disperse, while samples of grade B are quite concentrated to the centre, indicating a lesser spread of values. This is reflected in the sample standard deviation in Table 4. It might seem contradictory that a concentrated range of EMV values would lead to a large deviation; in fact, the opposite should be true. The explanation to this phenomenon is likely the fact that when heats were chosen, they were picked to ease examination of cross-sections; in other words, the heats with the most extreme values were picked, because that is where segregation was thought to be of maximum expanse. Furthermore, the deviations were calculated for a specific thickness, namely the thickness of 12mm and 15mm. This would further explain why the deviations does not add up with the figures. On the other hand, if the variable of thickness is disregarded, Figure 19 indicates that the samples in heats S1-S4 were generally more concentrated than heats H1-H4 shown in Figure 13. Thus, the EMV values of grade B can be interpret as being more uniform and consistent. In relation to segregation it is therefore possible to view grade B as the material where segregation plays the minor part. Hence, the segregation in grade A is believed to be of worse character and, as follows, indicates that centreline segregation is affecting impact toughness more than V-segregation. One possibility is that because the gradient in composition for grade A is more extreme and therefore could give more distinct extreme values in EMV.

5.7. Sources of error

The fact that specimens of type grade A showed a wider zone of columnar crystals compared to grade B, is likely to be in relation to theory discussed above. However, one important scenario to consider is that etching could affect how the surfaces is interpret by the human eye. Figure 8-11 indicates a straightforward appearance of columnar crystals in specimens of grade A, while almost none are distinct in Figure 14-17 of type grade B. The columnar crystals could very well be present in these samples as well, but no further investigation was made in this study.

Furthermore, there may exist other material defects that are unwanted and non-treatable in the test samples. For example, inclusions of oxides such as Al2O3, is a defect that cannot be treated by rolling or heating. Another example of defects that can affect the results are pores. These are especially important to consider in thick plate sheets where pores can travel deep into the material. In a thin material, the heat and force can reach the space around the pore and get rid of it. Therefore, the pores in a thick plate sheet can be present in the end of the production.

(27)

22

5.7.1. Measurement errors

Measurement errors are always present in every test and experiment. It is not known what errors exists in thickness measurements, and if the temperature was exactly −40°𝐶𝐶 in every Charpy-V test.

However, these minor factors are not believed to have had a significant effect on results. A major measurement error in this project is likely that 𝐶𝐶𝐶𝐶 is measured visually by an employee at the casting site. Even though the operators are trained and have references, it becomes somewhat subjective whether one sample should be categorized as 0 or 0,5 in 𝐶𝐶𝐶𝐶 index. Furthermore, it is not likely that the same operator has done all the inspections and judgements for the different heats included in this report. However, the importance is not in the actual 𝐶𝐶𝐶𝐶-value but rather that differences are visible.

For instance, a sample with 𝐶𝐶𝐶𝐶 = 0 visibly has a lower degree of segregation than a sample with 1 as 𝐶𝐶𝐶𝐶.

Another important aspect to cover is the variance in sample size across specimens of grade A and grade B. Unfortunately, the raw data received from SSAB was heavily biased in favour of grade B where at least 99,9980 % of the data was distributed. Of the selected data, corresponding percentage amounts to 87,18 %, which is also considered biased.

(28)

23

6. Conclusion

In short, this report has reached a likely conclusion that centreline segregation is affecting impact toughness more than V-segregation. Like discussed in previous chapter, specimens of grade A has shown to have a wider sample spread in general, ignoring the variable of sample size. In contrast, specimens of grade B illustrates a concentrated range of EMV values. From a customers point of view, a small spread of impact toughness is preferred as it ensures more homogenous properties in the product. The type of macro segregation in grade A is believed to be mostly made up of centreline segregation, while grade B is believed to have an increased tendency to form V-segregates. In accordance with the EMV-values it is therefore likely that macro segregation in the type of centreline segregates is worse than the type of V-segregates. This conclusion is of course ignoring possibly affecting variables such as temperature, composition, thermal contraction, solidification shrinkage, natural convection and equiaxed crystal growth; at least these are not measured by experiments.

However, the literature study supports the theory that centreline segregation is worse, and moreover it can be confirmed by the data when larger sample sizes are used. Thus, the conclusion that centreline segregation is the worst in relation to impact toughness is the most plausible.

One fact that supports this conclusion is the variable of 𝐶𝐶𝐶𝐶. It is shown in Table 2 and Table 4 that grade A in general seems to have a higher 𝐶𝐶𝐶𝐶-value than grade B, at least for the heats observed in this report. However, for the heats of grade A and B that had the same thickness and same 𝐶𝐶𝐶𝐶 value, the standard deviation was higher for grade B. This would lead to the thought that the segregation in grade B is affecting the impact toughness more since the samples otherwise have the same variables.

However, as stated in the discussion, 𝐶𝐶𝐶𝐶 is not a waterproof value in a scientific point of view since it is evaluated from visual inspection by an operator. Also stated in the discussion is how the sample size affects the analysis of data and that these effects can return biased results. Nevertheless, the general conclusion that centreline segregation is worse is still acceptable and highly plausible. Furthermore, most cases found in the literature study seems to agree with this conclusion.

(29)

24

7. Recommendations

For future research, an interesting aspect to consider could be temperature. It is common knowledge that most materials have transition temperatures where the material goes from e.g. ductile to brittle.

In this report, since a variety of different variables is likely to have a significant effect on the EMV value, temperature was one of the variables we could not consider, as the subject is too vast for this thesis report. It should be emphasised though, that all the Charpy-V tests were carried out under identical temperature, thereby eliminating at least some likely errors in measurements. Furthermore, one can find this exact temperature of -40°C listed in SSAB’s official product specification, across many different steel types; it seems to be standardized to test impact toughness at the temperature -40°C.

Since different materials are likely to have variations in both crystal- and lattice structure, it is especially interesting to consider the temperature usage of a material when designing the product.

One possible method to conduct this future research could be the usage of convolutional neural networks and deep learning through the process of grey scaling cross-section captures and putting them into the algorithm.

Originally, a test with laser on a cross-section of material was planned to be used to determine concentration deviations and thereby segregation degree in this project. This was cancelled because of time limitation but is something to recommend future studies on this subject because numbers are preferred on every phenomenon to easier compare and make conclusions. If the laser test had been conducted, a certain segregation degree would have been acquired, instead of only 𝐶𝐶𝐶𝐶-values. This could have improved the reliability of the conclusions.

(30)

25

8. Acknowledgement

This project was supported by and made in cooperation with SSAB Oxelösund. More specifically the continuous casting site. We would like to thank Fredrik Larsson, Process developer metallurgy SSAB Special Steels for the opportunity to visit the steel plant and setting up contacts that have been useful in our research.

We would also like to thank Pouyan Pirouznia, Process developer metallurgy SSAB Special Steels for sharing knowledge about macro segregations in the strand and regarding the continuous casting process. The answers to our questions that they provided greatly improved the quality of this paper.

Finally, we would like to thank our supervisor Anders Eliasson at Material science and engineering (MSE) at KTH, for supervising and checking in on our progress continuously throughout these past months and giving feedback.

(31)

26

9. References

[1] H. &. Å. U. Fredriksson, “Materials Processing during Casting, chapter 11,” John Wiley & Sons, Chichester, UK, 2006.

[2] V. 1. -. C. p. 3. ASM Handbook, ASM International, 2008; 2010.

[3] H. Fredriksson, “On the Solidification of Steel Ingots and Continuously Cast Steel Billets and Slabs, p. 235-244,” Canadian Metallurgical Quarterl, 1991.

[4] p. 1. H. &. Å. U. Fredriksson Materials Processing during Casting, “ Materials Processing during Casting p.164,” John Wiley & Sons, Chichester, UK, 2006.

[5] C. S.K, G. S, S. A and S. V, “Solidification morphology and segregation in continuously cast steel slab, p.312-321,” Journal of Materials Processing Technology, 2017.

[6] E. B. J. &. K. R. DeGarmo, “Materials and processes in manufacturing (7.th ed), p.285,”

Macmillan, New York, 1988.

[7] Jernkontoret, “Järn- och stålframställning Skänkmetallurgi och gjutning,” Jernkontoret, Stockholm, 2000.

[8] J. Priv. Doz. Dr. rer. nat. Dr.-Ing. habil. Hatto and W. Dr.-Ing. Klaus, “Final solidification, centre segregation and precipitation phenomena in continuously cast round billets, p.259-264,”

Verlag Stahleisen GmbH, 2016.

[9] T. Xu, “Detection of Final Solidification Due to Variation of Ferrostatic Pressure during Continuous Casting, p.2-7,” Stockholm, 2017.

[10] 浩. &. I. K. 磯部, “ブルーム連続鋳造における各種偏析の生成機構,” Tetsu-to-Hagane, 98(8), 405-414.. , 2012.

[11] 是. 宮. 剛. S. K. &. M. T. 鈴木, “鋼塊V偏析の生成機構について,” Tetsu-to-Hagane, 59(3), 431-445.. ) , 1973.

[12] G. Lesoult, “Macrosegregation in steel strands and ingots: Characterisation, formation and consequences,” Materials Science & Engineering A, 413 414, 19., 2005.

[13] A. Ghosh, “Segregation in cast products,” Sadhana, vol. 26, pp. 5-24, February 2001.

[14] W. K. Jacobi H, “Final solidification, centre segregation and precipitation phenomena in continuously cast round billets,” Steel Research. , Vols. Jun;68(6):258-65, 1997.

[15] B. V. R. J. Mihály Réger, Artist, Control of Centerline Segregation in Slab. [Art]. Acta Polytechnica Hungarica, 2014.

[16] M. B. V. R. J. Réger, “Control of centerline segregation in slab casting, p.119-137,” Acta Polytech. , vol. Hung11 , 2014.

[17] M. Seah, “Quantitative prediction of surface segregation, p.450-457,” Journal of Catalysis, vol.

57, no. 3, pp. 450-457, 1979.

[18] International Organization for Standardization, “ISO 148-1: Metallic materials, Charpy pendulum impact test,” International Organization for Standardization, 2016.

[19] W. Commons, “Wikipedia,” [Online]. Available:

https://commons.wikimedia.org/wiki/File:Charpy_V-notch_test.svg. [Accessed 03 12 2018].

(32)

27

[20] E. J. E. G. G. J. H. L. Blom G, Sannolikhetsteori och statistikteori med tillämpningar, p.370, 228, Studentlitteratur, 2017.

(33)

TRITA ITM-EX 2019:514

www.kth.se

Macro segregation in continuous cast HSLA steels